Article

"Self-Similarity and Compositional Strategies in the Music of Milton Babbitt"

John Cuciurean
Canadian University Music Review / Revue de musique des universités canadiennes, vol. 17, n° 2, 1997, p. 1-
16.

Pour citer cet article, utiliser l'information suivante :

URI: http://id.erudit.org/iderudit/1014783ar

DOI: 10.7202/1014783ar

Note : les règles d'écriture des références bibliographiques peuvent varier selon les différents domaines du savoir.

Ce document est protégé par la loi sur le droit d'auteur. L'utilisation des services d'Érudit (y compris la reproduction) est assujettie à sa politique

d'utilisation que vous pouvez consulter à l'URI https://apropos.erudit.org/fr/usagers/politique-dutilisation/

Érudit est un consortium interuniversitaire sans but lucratif composé de l'Université de Montréal, l'Université Laval et l'Université du Québec à
Montréal. Il a pour mission la promotion et la valorisation de la recherche. Érudit offre des services d'édition numérique de documents

scientifiques depuis 1998.

Pour communiquer avec les responsables d'Érudit : info@erudit.org

Document téléchargé le 30 September 2016 07:46

SELF-SIMILARITY AND COMPOSITIONAL
STRATEGIES IN THE MUSIC OF MILTON
BABBITT
John Cuciurean1

There would seem to be a justifiable expectation that composition in the

twelve-tone system would employ explicitly formalistic procedures to a
greater extent than triadic composition, not only because of the closed,
symmetrical nature of the pre-compositional materials of the system, and the
fact that the unit of definition (the set) defines a unique set of relationships
for each work, but, above all, because of the lack, in the twelve-tone system,
of the procedures of functional harmony. ... Thus, instead of a formalistic
result, twelve-tone composition would seem to require a predetermined
formalistic means.2 (Milton Babbitt, 1946)

The preceding quotation, taken from the fifth section of his doctoral disserta-
tion, reveals Babbitt's philosophy for formal implications of set structure in
twelve-tone music. In the ensuing forty pages, the thirty-year-old composer
intuitively described a compositional approach to large-scale form that would
guide his musical thought for the next half century. Babbitt's first works to
incorporate the compositional strategies outlined in his dissertation were
entitled Three Compositions for Piano (1947), Composition for Four Instru-
ments (1948), Composition for Twelve Instruments (1948/revised 1954), and
Composition for Viola (1950). In view of Babbitt's attitude towards how
mathematics effects twelve-tone procedures, the term "composition" in the
titles of these early works probably is meant to suggest its algebraic meaning,
that is, as one of the fundamental concepts of group theory.3

1A version of this paper was presented at the Canadian University Music Society's annual meeting
on 31 May 1996 at Brock University, St. Catharines, Ontario, Canada. I would like to convey my
appreciation to Martha Hyde, Jonathan Kochavi, Catherine Nolan, and Andrew Mead for reading through
earlier drafts of this paper and providing numerous insightful comments and helpful suggestions.
2Milton Babbitt, "The Function of Set Structure in the Twelve-Tone System" (Ph.D. diss.,
Princeton University, 1946/1992), 153-54.
3Since these first works, Babbitt has used "composition" in the titles of three additional works:
Composition for Tenor and Six Instruments (1960); Composition for Synthesizer (1961); and Composi-
tion for Guitar (1984). Always quick with a pun, Babbitt's titles frequently evoke multiple meanings
which often have humorous associations. For instance, Semi-simple Variations (1956), Sextets (1966)
and The Joy of More Sextets (1986) both for violin and piano, My Complements to Roger (1978), About
Time (1982), Four Play for four players (1984), It Takes Twelve to Tango (1984), Whirled Series (1987),
or Around the Horn for solo horn (1993) to cite but a handful. The multiplicity of meaning inherent in
his titles is carried over into his compositional approach.
2 CUMR/RMUC

In a number of published interviews, Babbitt, who presents himself as a

self-confessed maximalist, describes his aesthetic belief that just as in tonal
music, every musical event in post-tonal music should have multiple functions.
Beginning with his early works, Babbitt's compositional strategies reveal rich
and complex formal procedures, yet his theoretical writings on serial music are
frequently elliptical and fail to reveal a clear foundation for investigating the
deeper structural levels that form his music. On the other hand, the analyses of
individual pieces by scholars such as Joseph Dubiel, William Lake, David
Lewin, Andrew Mead, Robert Morris and Brian Alegant, and John Peel and
Cheryl Cramer have gone far towards establishing a foundation for an informed
theory of deep-level structure.4 These analyses as a group provide a basis for
this paper's topic, which is to explore in Babbitt's music the relationship
between the group properties of set structure and implied musical form at
deeper structural levels.
In "Twelve-Tone Invariants as Compositional Determinants," Babbitt out-
lines the mathematical group properties of the four transformational operators
P, I, R, RI.5 I would like to review the concepts of group theory that are
pertinent to my paper before discussing Babbitt's music. My reasons for this
will become clear during the course of my discussion. (For a more detailed
summary of my notational conventions, including the formal definitions of
group theory and analytic terminology used throughout the essay, please refer
to the appendix at the end of this article.)
Briefly, a mathematical group consists of a non-empty set of abstract
objects, and a binary operator defined on the set of objects which have four
properties: closure, associativity, identity, and inverse. Our usual model of
pitch-class space (henceforth, pc space) is a cyclic group Z12 which consists of
the set of the integers {0, 1,..., 11} together with the binary operation of mod
12 addition. The direct product of the dihedral group D2, which is comprised
of the four classical serial operators {T, I, R, RI}, and the cyclic group Z12
which models pc space produces the forty-eight canonical operators — a direct
product group which yields a homomorphic image of Z12xD2. (See definitions
2, 3, and 4 in the appendix.)
The parallel relationship shared between the structural characteristics of a
parent group and one of its semi-groups is the essence of "self-similarity." The

4 Joseph Dubiel, "Three Essays on Milton Babbitt - Part One: Introduction, Thick Array of Depth
Immeasurable," Perspectives of New Music 28, no. 2 (1990): 216-61; William Lake, 'The Architecture
of a Superarray Composition: Milton Babbitt's String Quartet No.5" Perspectives of New Music 24,
no. 2 (1986): 88-111; David Lewin, "Generalized Interval Systems for Babbitt's Lists and for Schoen-
berg's String Trio," Music Theory Spectrum 17, no. 1 (1995): 81-118; Andrew Mead, "Detail and the
Array in Milton Babbitt's My Complements to Roger" Music Theory Spectrum 5 (1983): 89-109;
Andrew Mead, "Recent Developments in the Music of Milton Babbitt," Musical Quarterly 70, no. 3
(1984): 310-31; Andrew Mead, An Introduction to the Music of Milton Babbitt (NJ: Princeton
University Press, 1994); Robert Morris and Brian Alegant, "The Even Partitions in Twelve-Tone
Music," Music Theory Spectrum 10 (1988): 74-101; and John Peel and Cheryl Cramer, "Correspon-
dences and Associations in Milton Babbitt's Reflections" Perspectives of New Music 24, no. 2 (1986):
144-207.
5 Milton Babbitt, "Twelve-Tone Invariants as Compositional Determinants," The Musical Quar-
terly 46, no. 2 (1960): 246-59.
17/2(1997) 3

concepts of self-similarity are used in mathematics and physics to describe the

geometry of crystalline structures and serve as the basis for fractal imaging,
yet few music theorists have explored self-similarity as a basis for structural-
level analysis of musical form.6 The concept of self-similarity is one manife-
station of the subset invariance theorem. Applying this theorem to a subgroup
of the four (of forty-eight) canonical operators that map a type A, B, C, or E7
hexachordally combinatorial row form to a row form which maintains hexa-
chordal-invariance produces the results shown in table l: 8

TQ In+6 ^6 RIn

T0 ~TÔ ln+6 R6 R^
In+6 In+6 T0 RI n R6

ks R6 RI n T0 In+6

k RIn R6 In+6 T0 J

Table 1: The direct product of the four classical serial operators, T, I, R, RI.

{T0, In+6, R6, RIn) = D2 is a subgroup of Z12 x D2, which holds hexachordal pc
content invariant for type A hexachords at n=l 1, type B at n=l, type C at n=3
and type E at n=3, 7, ll. 9 (Also note that if we wish to examine hexachordal
combinatorial relationships rather than hexachordal pc invariance relation-
ships, all that is required is to re-map the subgroup {T0, In+6, R6, RIn} via the

6Some may argue that Schenker's notion of motivic parallelism (as discussed in Charles Burkhart,
"Schenker's 'Motivic Parallelisms'," Journal ofMusic Theory 22, no. 2 (1978): 145-75; and elsewhere)
on the various structural levels — especially between surface gestures and the Urlinie — constitute
analysis of self-similar properties. Analogously, I use the English terms foreground, middleground, and
background when discussing Babbitt's structural levels, but my analysis focuses on direct mathematical
isomorphisms and it is at this juncture that the philosophical basis of Schenker's exploration of
self-similar relationships and mine diverge.
7Type A, B, C, and E hexachords are defined by Babbitt in "Some Aspects of Twelve-Tone
Composition," The Score and LM.A. Magazine 12 (1955): 53-61 as four of the six possible all-
combinatorial hexachords; the other two are designated types D and F. For readers more comfortable
with Forte set-class labels, types A, B, C, D, E, and F hexachords are equivalent to set-classes 6-1
(012345), 6-8 (023457), 6-32 (024579), 6-7 (012678), 6-20 (014589), and 6-35 (02468t) respectively
as given in Allen Forte, The Structure of Atonal Music (New Haven: Yale University Press, 1973).
Initially I exclude D and F type hexachordsfromthe subset invariance theorem due to the complications
introduced by the tritone symmetry inherent in these two hexachords. I will develop and expand the
subset invariance theorem latter in this paper to include the type D hexachord. I will not expand the
theorem to include the type F "whole-tone" hexachord in this paper since Babbitt avoids this hexachord
in his compositional practice.
81 am grateful to Larry Fritts for bringing this example to my attention in his paper "The Group
Structure of Babbitt's Three Compositions for Piano** read at the Music Theory Midwest/Society of
Composers Meeting, University of Iowa, 7 April 1995. Discrepancies between our two presentations
are a result of different notational approaches to the same problem.
9 Note that each of these four hexachord types contain a zero entry for interval class 6 in the ic
vector which accounts for the Ré invariance of each. Also note that these n values apply to sets in prime
form. The interval-class vector for a type A hexachord is <543210>; type B hexachord is <343230>;
type C hexachord is <143250>; and type E hexachords is <303630>.
4 CUMR/RMUC

T6 operator onto a coset {T6, In, RQ, RIn+6}, where the respective values of n
remain unchanged for each of the four represented hexachord types.)
I propose that the generation of cosets of the forty-eight canonical operators
whose membership is contingent upon the hexachordal combinatorial rela-
tionships of a row form provides the basis for large-scale structure in Babbitt's
music. In the following section I examine how Babbitt selects and combines
rows to produce maximal diversity on the surface while insuring maximal
internal coherence at the deeper structural levels. My examination focuses
exclusively on pitch space, but I see little reason why this same analytic
principle could not be extended to the rhythmic domain — indeed, the recent
work by Andrew Mead supports this contention by revealing parallels that exist
between Babbitt's pitch and rhythmic structures.10 My investigation took
several of Babbitt's pieces into consideration, spanning almost fifty years of
compositional practice. In the interest of space, I concentrate on three pieces,
Duet for Solo Piano (1956), Reflections for Piano and Tape (1975), and Soli
e Duettinifor Two Guitars (1989) which together use straightforward serial
structures, array structures, and superarray structures respectively.
Babbitt's technical procedures are gradually becoming documented through
articles that address both analytical and theoretical issues.11 Most importantly,
several published analyses (including those by Dubiel, Lake, Mead, and Peel and
Cramer) have focused on the difficult issue of how the composer incorporates
pitch-array strategies. My paper investigates the structure of pitch arrays and
demonstrates how the hexachordally combinatorial properties of a given row can
be used as transformational operators on the structural middleground to produce
a complex network of row forms, all of which relate back to the dihedral group
properties of the four transformational operators at the background structural-
level, while maintaining maximal diversity on the foreground.

Simple Serial Structures

Duet for Solo Piano (1956), which was composed as a gift to his daughter,
concisely illustrates Babbitt's strategy for large-scale structural design. The
polyphony is based exclusively on a two-voice counterpoint between the
pianists right and left hands. The initial row form, <20795463t8el>, presented
in the right hand beginning at m. 1 derives from a type C all-combinatorial
hexachord which is hexachordally combinatorial when integrated with the
coset of canonical operators, {T6,13, Ro, RI9} : refer back to Table 1. The piece
uses the following progression of row forms on its surface or foreground:
Measure no.: 1 6 10 14
Lyne 1 (RH): P^ RI^ h, R^
Lyne2(LH): P8 R2 RI 7 Ii
Example 1: Row form array for Duet

10Andrew Mead, "About About Time's Time: A Survey of Milton Babbitt's Recent Rhythmic
Practice," Perspectives of New Music 25, nos. 1-2 (1987): 182-235.
11 As the list of citations in footnote no. 4 suggests.
17/2(1997) 5

We first need to examine the relationships between the eight row forms that
structure the piece. The eight row forms, {P2, RIlf I7, Rg, Pg, R2, RI7, Ii}, form
a collection which is generated from a single prime-form set P2, its T0 trans-
position (trivially), its I3 inversion, their T6 transpositions and the RQ retro-
grades of all sets derived thus far, shown in equation 1.
Eq. 1 {P2}®{T0, T6,13,19, Ro, Re, RI3, RI9}={P2, Pg, Ii, I7, R2, Rg, Wi, RI?}
® denotes an operator whose operands consist of a set of row forms and a set
of canonical operators. The operation produces a resultant set of row forms
which is comprised of the union of the application of each and every member
of the canonical operator set to each and every member of the row form set, in
no specific order.
To generalize this from a slightly different perspective, I will call P8 the
hexachordal complement of P2, and label P 8 sP 2 -l, then equation 2, which
includes {P2, P 2 ~l} as the set of row forms and {T6,13, RQ, RI9} as the coset
of canonical operators which effect hexachordal combinatoriality for a type C
all-combinatorial hexachord, also describes the pc content of this piece.
Eq. 2 {P2, P2~l}<8>{T6,13, Ro, RI9}s{P2, Pg, Ilf I7, R2, Rg, Rib RI?}
This equation could also be expressed using the subgroup of operators which
hold hexachordal content invariant, shown in equation 3.
Eq. 3 {P2, P2~1}®{T0,19, R6, RI3MP2, Pg, Ii, I7, R2> Rg, Wi, RI7}
Please note that equations 1, 2, and 3 are all equivalent. Furthermore, in each
case I chose P2 as my reference row form, but I could have used any one of the
eight representative row forms in the piece as my reference point and produced
the same 8 row collection in each case.
Returning to the array of row forms distributed throughout the piece, another
important property structures the horizontal relationships between the succes-
sive row forms in each hand and the vertical relationships of the simultaneously
presented adjacent row forms. The diagram in Example 2 presents a chronolo-
gical account of the progression of row forms reading left to right and an
account of the combinatorially paired row forms reading top to bottom. The
large arches that span from start to finish account for the net result of the
composite linear operations.12
As we might expect, the vertical pairings of row forms consistently take
advantage of the coset of combinatorial operators, {T6,13, RQ, RI 9 }. Moreover,
and in keeping with Babbitt's edict of maximal diversity, each one of the four
operators is used one time only. What seems less predictable, however, is the
set of operators acting upon the linear or horizontal transformations of succes-
sive row forms. The set of operators that move the piece temporally forward
12The network presentation is modelled after Henry Klumpenhower, "A Generalized Model of
Voice-leading for Atonal Music" (Ph.D. diss., Harvard University, 1991); David Lewin, Generalized
Musical Intervals and Transformations (New Haven: Yale University Press, 1987); and David Lewin,
"Klumpenhower Networks and Some Isographies That Involve Them," Music Theory Spectrum 12, no.
1 (1990): 83-120.
Example 2: Transformational network of pitch structure in Duet.

consist of {I9, R$, RI3}, which if we admit the identity operator T0 to this set,
constitutes the subgroup of operators that maintain pc hexachordal invariance
for a type C hexachord, {T0,19, R6, RI 3 }.
Just as we created a complement relation, {P2, P2"1}» between two distinct
row forms, we can define the complement-set of canonical operators to the
coset of hexachordally combinatorial operators as abstractly being the sub-
group of operators which maintain hexachordal pc invariance. (Hie formal
definition for this equation is included in the appendix, Definition 6.) If Ss{T6,
In, RQ, RIn+6}, a coset of D2xZ12 that admits hexachordal combinatoriality for
type C hexachords at n=3, then we shall recognize S"1 as given in equation 4
to be a subgroup of D2xZ12 that holds hexachordal pc content invariant for type
C hexachords at n=3.

We can now abstractly identify the background structure of this brief

composition. By background structure, I simply mean that the entire piece can
be represented as the direct product of a single, well defined set of row
forms and a subgroup of operators. By beginning with the pitch map at the
foreground, shown in Example 3, the algebraic model is developed as the
middle- and background levels are successively extracted from the fore-
ground.
The analytic model in Example 3 marries algebraic self-similarity with
multi-level structural paradigms, yielding what I believe to be a novel, albeit
complex, perspective on an otherwise simple serial piece. Some critics might
argue that this analytical approach is comparable to tapping a thumb tack into
the wall with a sledge hammer, and perhaps in this simple case such an
 17/2(1997)

argument has some merit. However, its application to Reflections for Piano
and Tape (1975) suggests that the proverbial thumb tack has developed into a
large spike and the analytic apparatus no longer seems quite so extravagant.
For convenience, I here use John Peel and Cheryl Cramer's analysis of
Reflections as my point of departure.13

Array Structures
The all-combinatorial properties of the hexachord that structures Reflec-
tions' type A row are exploited in the construction of the 12-lyne, 6-block,
77-partition, all-partition array which forms the foundation for the piece's
pitch structure.14 The transformational network which generates the fore-
ground pitch-class all-partition array in the first section of the piece (mm. 1-82)
is shown in Table 2.

13 Peel and Cramer, "Correspondences and Associations." I draw primarily upon the chart of row
forms that Peel and Cramer present as structuring the surface of the form, given on pp. 186-203. Note
that wherever they have identified a prime row form as Sn, I have rewritten the label as Pn.
14Ibid. Babbitt himself first outlined partition array strategies in "Since Schoenberg," Perspectives
of New Music 12, nos. 1-2 (1974): 3-28, however I refer the reader to the recent writings of Andrew
Mead and William Lake for a clearer explication of partition arrays. See for example Andrew Mead,
"Detail and the Array"; idem, An Introduction to the Music of Milton Babbitt (Princeton, 1994),
124-203; William Lake, "The Architecture of a Superarray."
g CUMR/RMUC

Blackl Block 2 Block 3 Block 4 BlQ£ki BiQCli

W7 I. R« P2 RI7 I
1 I» R2 P8 RI, I7 R2 J
RI, I3 Rio P4 RI9 1
1 I. R4 P.o RI3 I» R4 /
Po RI5 In R« Po RI5 I
( RI„ I5 Ro J
Ro P«

| Rl 2 P9 R3 I, Rl2 P, I
1 R, I2 RI, p3 R, I2 /
| RI4 p., R5 I.o RI4 P„ 1
l Ru I4 RI.o P5 Ru I4 J
I Rio P7 R. 16 Rio
I R7 Io RIe P, R7 Io /

Table 2: Row form array for Reflections (mm. 1-82).

By the end of the fourth block of the array, all forty-eight forms of the row
have appeared once and only once. Blocks 5 and 6 use row forms which have
previously appeared in blocks 1 and 2 exclusively. Careful analysis of this
network reveals that the lynes of the array are grouped by row pairs that under
I5 are hexachordally combinatorial (indicated with curly brace brackets). These
are further related by treating the lower six lynes as an RI2 transform of the
upper six lynes. To carry this network of transformational relationships one
level deeper I have labelled the twelve row forms that make up array block 1
as Al and I have traced the progression of row forms temporally through the
array. This resultant network of horizontal transformations is shown in the
middleground analysis of Example 4.

Example 4: Middleground network of array structure in Reflections (mm. 1-82).

Returning to the foreground in the second section of the piece (mm. 8 3 -

178), the pc array utilizes a different distribution of row forms, as shown
in Table 3. The use of all forty-eight row forms by the end of the fourth
block is still in place, as are the row pairs that are I5 hexachordally
combinatorial (indicated with curly brace brackets) and the RI2 relationship
between upper and lower six lynes.
17/2(1997) 9

Block 1 BjQ£k_2 Bteskl Block 4 BIQCJLS Block 6

Ro Is Mu p6 Ro 15
RI5 RI5 )
Po R. III Po J
RI, Pio R* I, RI3 Pio I
Rio I3 RI9 P4 Rio I3 /
RI, p, R2 I7 RI,
R, I. RI7 p2 R, I. J

I» R3 P, RIj I. R, J
P3 RI8 I2 R, P3 RI, J
Iio R5 P.. RI4 1.0 R5 1
P5 W10 I* R11 P5 RI10/
I* Ri P7 RI» I6 R, I
P, w. Io R7 p. RI6 /

Table 3: Row form array for Reflections (mm. 83-178).

Returning to the middleground analysis, I have labelled the twelve row forms
that comprise block 1 as A2 for this second section of the piece. The resulting
network of horizontal transformations is shown in the middleground analysis
given in Example 5. Please take a moment to note the homomorphism that
characterizes the relationship between the transformational networks of the
first and second sections of this work by comparing Examples 4 and 5.

Example 5: Middleground network of array structure in Reflections (mm. 83-178).

To reduce the structure one more level, I shall denote §1 and §2 to represent
the pc arrays for sections 1 and 2 of the piece respectively, as given in Equations
5 and 6:

From m. 179 to the end of the piece, Babbitt juxtaposes 111 transformations of
ordered structures §1 and §2, which yields the formal structure shown in
Example 6.

Example 6: Middleground network of array structure in Reflections.

10 CUMIURMUC

The transformations employed at the middleground level are once again

restricted to the coset of four hexachordal-combinatorial operators and the
subgroup of four hexachordal-invariance operators for type A hexachords,
S-1={T0,15, R6, RI n } and S={T6, I u , R0, RI5} respectively. As in Duet, this
achieves compositional unity on the middleground structure while simulta-
neously exercising the concept of maximal diversity on the foreground; in this
piece Babbitt manages to circulate through all forty-eight row forms four times
at the foreground level. Moving toward the background, the algebraic model I
offer for the structure of Reflections is provided in Example 7.

Example 7: Analytic model of Reflections at successive reductional levels.

At this point the need for levels of abstraction begins to come into better
focus, for if I were to attempt to interpolate this middleground structure into a
shallower middleground and eventually to the foreground structure — which
was originally provided in the Peel and Cramer analysis — the level of
complexity would render the model ineffective.15
Further analysis of the pitch structure in Reflections concerning the inter-
relatedness of individual row forms within Al and A2 or between Al and A2
have yielded inconclusive results, or more correctly, have generated subgroups
of operators that do not allow for the tidy analytical models based on the subset
invariance theorem that I have been developing thus far. This does not mean
that these relationships are uninteresting, but since they are tangential to the
correlation between pitch structure and the subset invariance theorem, I need
not explore them further here.

151 need to begin with the foreground analysis in order to extract the middle- and background
structures — my analytical methodology must start with the surface pc's and unidirectionally move
toward the background algebraic model.
17/2(1997) 11

Superarray Structures
Soli e Duettinifor Two Guitars (1989), the first in a series of three pieces that
all share the title Soli e Duettini, is an example of Babbitt's most recent
compositional practice of using nested array structures to produce what several
scholars have described as superarray structures.16 To date, the analytic litera-
ture for these three particular pieces is confined to a brief discussion of their
large-scale superarray design in the closing pages of the final chapter in
Andrew Mead's recent study of Babbitt's music.17
The twelve-tone row for the piece for two guitars, <0546te293187>, derives
from a type D, second order all-combinatorial hexachord which is hexachor-
dally combinatorial when integrated with the coset of canonical operators, {T3,
T 9 ,1 5 , In, Ro, ^6» RI2, Rig} and hexachordally invariant when combined with
the subgroup of canonical operators {T0, T 6 ,1 2 ,1 8 , R3, R9, RI5, RIn }.18
The basic array for Soli e Duettini is based on a 6-lyne, 8-block, 58-partition
array which incorporates each one of the 58 unique partitioning patterns of 6
lynes as well as all 48 distinct forms of the twelve-tone row once and only
once.19 The basic array for guitar 1 is a T6 transformation of the array used
for the violin part in The Joy of More Sextets while the basic array for guitar
2 is an M7R4 transformation of the basic array used for guitar l.20 Table 4
provides the transformation network which structures the foreground pc array
for guitar 1.

16The other two pieces are Soli e Duettinifor Guitar and Flute (1989) and Soli e Duettinifor Violin
and Viola (1990). In An Introduction to the Music of Milton Babbitt, 204, Andrew Mead points out that
many of the pieces from Babbitt's third period, beginning with Ars Combinatoria (1981), incorporate
superarray strategies.
17Ibid., 255-63.
18 Recall from the discussion of the subset invariance theorem in the first section of this paper that
a type D hexachord is one of the six all combinatorial hexachords defined by Babbitt. Specifically, this
type is a second-order, all-combinatorial hexachord which has the Forte set-class label 6-7, its prime
form is (012678) and its interval-class vector is <420243>. Notice that the value for n in the subset
invariance theorem is similar to that of the type A hexachord used in Reflections with some modifications
to accommodate the tritone symmetry inherent in the second-order all-combinatorial type D hexachord.
To begin let n=l 1, as is the case in the subgroup for the type A hexachord from the original theorem
(see Table 1), but modify the subgroup to read {To, In+3, R3, RIn}. Next, combine the subgroup with the
coset {TO, In+9, R9, RIn+6}, produced via a T6 transformation on the subgroup itself, yielding the total
hexachordal invariance subgroup S'^fTo, Té, In+3, In+9, R3, R9 RIn, RIn+6}. Re-mapping the hexachor-
dal pc invariance subgroup via the T3 operator onto the coset, S={T3, T9, In, In+6, Ro, Ro, RIn+3, RIn+9},
will produce the hexachordal combinatorial operators for the type D hexachord. These modifications
account for the difference between the type A hexachord which is accurately described as a 6-pc
chromatic cluster and the type D hexachord which may be best characterized as two disjunct 3-pc
chromatic clusters a tritone apart.
19Ibid., 271. Mead describes the array for Soli e Duettini as being the same form as the array used
for the solo violin part in The Joy of More Sextets, which is provided on pages 278-79.
20M7 is the multiplicative operator which maps a given row form onto its circle-of-fifths transform.
To generalize the multiplicative operator using the same format as the generalized models of the four
canonical operators: Let X be the set of all rows; Let Xy=(xi, X2,..., xn)€ X; Define the transformation
onX: Mn:X-»X, whereM n (X y )=(nxi,n-X2,..., n x n ) mod 12.
 12 CUMR/RMUC

Blûfitl moskl mskl moskà moskl SIÛCLÉ BlfifiLZ Btocki

J It W* P4 Ri Pio RIii h R7 1
p
l ? R4 1,1 RI8 l5 R, 0 P. Rl 2 J
I Ï9 Pu R2 Rio R8 P5 l3 Rl 6 1
I R5 RI3 io P2 l6 RI 9 R„ P« J
. J ** I7 R3 RI4 P9 Ii Ro RI,o 1
p
I 6 RI7 R9 I» R3 RI, P0 l4 /

Table 4: Row form array for guitar 1 in Soli e Duettini.

As in our previous analysis, the lynes of the array are grouped into hexa-
chordally combinatorial pairs. Unlike the lyne pairs in the array used in
Reflections which all shared the same transformational operator for their
combinatorial pairings, each of the lyne pairs in this array has a unique transfor-
mational operator to effect hexachordal combinatoriality. The upper lyne pairs are
related by I5 or I n , the middle lyne pairs by RI2 or RI8, and the lower lyne pairs
by RQ or R6. If we group these six operators in a set along with T3 and T9, which
are the two operators that allow us to map transpositionally the first hexachord
of the row onto the second, we produce the coset of hexachordally combinato-
rial canonical operators for a type D hexachord, S={T3, T 9 ,1 5 , I u , RQ, R6, RI2,
Rl 8 }.
Referring back to Table 4 for a moment, let the column of row forms which
comprise Block 1 be called A, such that

We can now follow the temporal progression of lyne pairs through the array at
a deeper structural level by inspecting the linear network of transformational
operators which operate on A, as shown in Example 8.

Example 8: Middleground network of simple-array structure in Soli e Duettini.

As was the case in the previous two analyses, the network of horizontal
transformations consists exclusively of the subgroup of canonical operators
which produce hexachordal invariance, S - ^ T Q , T 6 ,1 2 ,1 8 , R3, R9, RI5, R I H } .
Moving yet another level deeper in the structure of this piece, the partition-
ing strategy for the superarray is based on a series of alternating duets and solo
passages which are delineated by the eight blocks of the basic arrays for the
two guitars forming a ten-block superarray. Andrew Mead diagrams the super-
array structures for all three pieces titled Soli e Duettini and lines them up for
comparison.21 Following Mead, I have labelled the blocks in guitar l's basic

21 Mead, An Introduction to the Music of Milton Babbitt, 256.

17/2(1997) 13

array from 1 to 8 and the blocks in guitar 2's basic array from 8 to 1 to reinforce
the retrograde relationship between the two basic arrays as shown in Example
9. I have added the measure numbers and superarray block numbers above
Mead's diagram for clarity.

Example 9: Block analysis of superarray structure in Soli e Duetdni.

Blocks from the basic arrays marked with an asterisk contain an extra
vertical aggregate to accommodate the partitioning strategy, therefore Babbitt
is forced to work with unequal rates of aggregate completion between the two
guitar parts in the first, third, fourth and sixth duets (i.e., superarray blocks 1,
5, 7, and 10). In the second and fourth duets, Babbitt maintains a consistent
one to one aggregate correspondence between the two parts. By combining
Examples 8 and 9 we are able to algebraically model the middle- and back-
ground pitch structure as shown in Example 10 below.

In the middleground equation guitar 1 is the upper portion of the expression

and guitar 2 the lower. The dashed lines (-) in the shallow middleground
expression represent the places where the respective guitar parts are silent in
the ten-block superarray structure.
14 CUMR/RMUC

It should be evident at this point that the iterative quality of nested arrays
does not significantly alter the deep level structure as Babbitt moves from a
simple serial structure through to a superarray design. In fact, in each of the
three analyses in this paper the background model has been based on a single
row form or coset of row forms composed with the set theoretic union of the
transformational operators that produce hexachordal combinatoriality (S) and
hexachordal invariance (S-1). The main difference between the simple serial
structural model and the superarray structural model is the inclusion of the
transformational operator that generates the other forms of the basic array into
the equation.

Conclusion
Before I close, I would like to offer the following postulate: Given a twelve-
tone row based upon an all-combinatorial hexachord that lacks a tritone in its
ic-vector, the potential subgroup of operators that can generate maximal
diversity while maintaining maximum internal structural coherence would be
defined according to the subset invariance theorem (as worked out in Table 1)
and the resultant group would manifest itself on the surface through the
exploitation of hexachordally combinatorial pairings of row forms, even if the
partitioning strategy did not favour hexachords.
I will not attempt to prove this postulate here, but based on the analyses I
have completed thus far, I am persuaded that these are reasonable statements
to make. Moreover, I believe further queries along these lines will yield results
that would both strengthen and expand my current analytical model. Like many
others, I believe that the richly varied surface texture of Babbitt's work
distracts the listener from recognizing the inherent formal unities that structure
deeper levels of his music. I do not believe that this fact alone prevents the
informed listener or interpreter from detecting, at least subconsciously, the
underlying formal unities in this repertoire, but it certainly requires a reassess-
ment of the listening process and toward that end, new methods of dealing with
the surface of this music. More directly, the rich and varied textures that mark
the surface of Babbitt's twelve-tone compositional practice often seem to
obstruct or block simpler underlying structures. Nevertheless, just as we
tolerate and enjoy the complex strategies that underlie tonal structural levels,
we need to make an equally strong effort to come to grips with the intricacies
of Babbitt's extended serial compositions.
Appendix
Notational conventions
A,B mathematical sets
AuB union of sets A and B
a,b€A a and b are members of set A
{a,b} unordered set consisting of elements a and b
(a,b) ordered set consisting of a followed by b
a"1 inverse of a with respect to a given operation
a=b a is abstractly equivalent to b
a==b a approximately equivalent to b
17/2(1997) 15

Y: A-»B operation Y on A maps A onto B

A£l: A-»B the application of operation Cl to A and then A to the result of the first
operation onto B, called the composition of maps Q, and A.
2. Twelve-tone row forms
Xy=(xi, X2,..., X12), an ordered 12-tuple which exhibits the following properties:
i) the 12 entries of Xy consist of all the integers from 0 to 11 ;
ii) Xy€ {Py, Iy, Ry, RIy} where 0<y<l 1, and
Py Prime row form label beginning with pc y
Iy Inversion row form label beginning with pc y
Ry Retrograde row form label ending with pc y
RIy Retrograde Inversion row form label ending with pc y
3. Canonical operators
To define the four canonical operators (i.e., transposition (T), inversion (I),
retrograde (R), and the composite operator retrograde-inversion (RI)):
Let X be the set of all rows (i.e., all permutations of the integers 0 to 11);
Let Xy=(x!, x 2 ,..., x12) € X;
Define the following transformations on X:
Tn:X-»X, where Tn(Xy) = (x^n, x2+n,..., x12+n) mod 12
In:X-»X, where In(Xy) = (n-xj, n-x2,..., n-x12) mod 12
Rn:X->X, where Rn(Xy) =( x_! mod 13+n, x_2 mod B +n,..., x.12 mod 13+n) mod 12
= (x12+n, xn+n,..., Xi+n) mod 12
RIn:X->X, where RIn(Xy) = (n-x_! mod 13, n-x_2 mod 13,..., n-x_12 mod 13) mod 12
= (n-x12, n-x n ,..., n-xi) mod 12

4. Definitions
Definition l: 22 A mathematical group (G, °) consists of a non-empty set of abstract
objects, G, and a binary operator, °, defined on the elements of G which have the
following properties:
i) closure: x°ye G for all x,ye G;
ii) associativity: (x°y)°z=xo0(yoz) for all x,y,ze G;
iii) identity: there exists an element e€ G, such that x°e=*e °x=x, for all xe G;
iv) inverse: for all x€ G, there exists an x~*e G, such that x""lox=x°x-1=£.
Definition 2: ZQ is a cyclic group which consists of a set of n integers {0, 1,..., n-1},
together with the binary operation of mod n addition, such that the equation i+j=k
(mod n) is always satisfied for any 2 integers i, j<n and their integer sum k (mod n)<n.
Definition 3: Dm is a dihedral group of order 2m which is isometric under rotation,
translation or reflection. The specific case of D2 can be thought of as the non-cyclic
group of order 4 which is generated by the cross product of Z2xZ2 which is a specific

22Based on definitions and theorems from H.A. Elliot, K.D. Fryer, J.C. Gardner and Norman J.
Hill, Vectors, Matrices and Algebraic Structures (Toronto: Holt, Rinehart and Winston, 1980), 367-407;
and Derek J. S. Robinson, A Course in the Theory of Groups, Graduate Texts in Mathematics, no. 80,
ed. J.H. Ewing, F.W. Gehring and P.R. Halmos (New York: Springer Verlag, 1991), 1-42.
16 CUMR/RMUC

direct product group known as the Klein-4 group in geometrical group theory (after
ninteenth-century mathematician Felix Klein).
Definition 4: DmxZn is a direct product group generated by the cross product of D m with
Zn, and has order 2mn.
Definition 5: Any given subset of a group which remains closed under the parent group's
binary operation is referred to as a semi-group. The semi-group will exhibit some of
the same structural characteristics as the parent group, but may lack either the identity
property or the inverse property or both properties of the parent group. If the subset
of the group also lacks the closure property under the parent group's binary operation,
the subset is referred to as a coset.
Definition 6: If S is defined as a coset of D2XZ12, and S represents the unordered set of
canonical operations that result in hexachordal combinatoriality for a given
twelve-tone row, then we shall recognize its inverse, S_1=T6(S)=D2, a subgroup of
D2XZ12 that represents the unordered set of canonical operations that result in
hexachordal invariance for the same twelve-tone row.
Subset Invariance Theorem: for any group G which permutes elements of some ordered
set U, there exists a non-trivial subgroup H which fixes the content of some non-trivial
subsegment S of U, while permuting its elements.
Definition 7: Lyne is analgous to a monophonie interpretation of a linear series of pc sets
or row forms.23 In Babbitt's music, lynes are usually distinguished by register, dynamic,
rhythmic pattern, articulation, orchestration, or any combination of the above.
Definition 8: An array is an abstract combination of two or more simultaneous horizontal
row forms or lynes which can be partitioned into columns that form vertical
aggregates.24 Babbitt incorporates several important classes of arrays into his
compositional language including trichordal arrays, all-partition arrays and
superarrays (i.e., arrays of arrays).25

Abstract
The unfolding of a compact algebraic group into a larger structure which exhibits an
isomorphic relationship with the smaller group is the essence of "self-similarity."
Through the use of transformational networks which take advantage of the group
properties of the forty-eight canonical operators and through the examination of the
hexachordally combinatorial properties of Babbitt's row forms, this paper examines the
manner in which Babbitt selects and combines rows to produce maximal diversity on the
surface while optimizing internal coherence at the deeper structural levels. This study
focuses on three works that cover straightforward serial structures, simple array structures
and superarray structures respectively — Babbitt's three main compositional strategies.

23 The term "lyne" is introduced by Michael Kassler in "Toward a Theory That Is the Twelve-
Note-Class System," Perspectives of New Music 5, no. 2 (1967): 1-80.
24The term "array" was first used by Godfrey Winham in "Composition with Arrays," Perspectives
of New Music 9, no. 1 (1970): 43-67.
25It is beyond the scope of this paper to provide genralized models of these array classes. The
interested reader can refer to Andrew Mead, An Introduction to the Music ofMilton Babbitt for a detailed
description of these various types of arrays and analyses of Babbitt's idiomatic strategies for
incorporating these arrays into his music. Mead loosely divides Babbitt's career into three creative
periods, each marked by an affinity for a particular array class: 1947-60, trichordal arrays; 1961-80,
all-partition arrays; 1981-present, superarrays.